Software development for the use of a mathematical prediction tool for chemotherapy responsiveness in a clinical and in a research environment

2019-11-22T18:34:44Z (GMT) by Elisabeth Zink

Since cancer is the second most common cause of death in Ireland (Irish Cancer Society, 2013), there is great potential to improve personal treatment of cancer patients and to reduce the mortality rate.

The search for biomarkers to predict the responsiveness to chemotherapy fails for certain cancer types due to their high heterogeneity. Integrating systems knowledge into biomarker research can improve the classification of patients into responder and non-responder. For instance the systems model APOPTO-CELL, which was developed in our group previously, predicts the treatment outcome of stage II and stage III colorectal cancer patients (Hector et al, 2012). This mathematical model outperforms classical statistical methods and is able to calculate whether targeted therapeutics such as apoptosis sensitizers may improve the responsiveness of individual patients to chemotherapy. Interviews with physicians and pathologists revealed that there is a need for prediction tools such as APOPTO-CELL that can support the consultant in the treatment decision process. APOPTO-CELLup could be validated successfully against the original published version. We therefore investigated how APOPTO-CELL may be integrated into a clinical diagnostic environment. For this purpose a requirement analysis was performed and as a result two possible use cases, one in the clinical pathology and the other in the clinical research setting were defined. Based on the use cases, a prototype for a graphical user interface was developed. Feedback from potential users resulted in a refined user interface that can be integrated into existing clinical workflows. APOPTO-CELL provides the prediction of treatment responsiveness in different standardized exchange formats. Here, I present a workflow integration strategy and a functional prototype for the APOPTO-CELL response prediction. A final application of the model to a patient set with 221 stage II and III CRC patients revealed that the model is able to predict chemotherapy responsiveness in patients with stage III CRC and classify correctly responders and non- responders to chemotherapy.